2015 Informs Annual Meeting

SD44

INFORMS Philadelphia – 2015

SD45 45-Room 103C, CC Pricing — Examples of Collaboration Between Academia and Industry Sponsor: Revenue Management and Pricing Sponsored Session Chair: Georgia Perakis, MIT, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States of America, georgiap@mit.edu 1 - Balancing Profit Maximization and Inventory for Recommending Personalized Bundles Anna Papush, MIT, 77 Massachusetts Avenue, Cambridge, MA, United States of America, apapush@mit.edu, Georgia Perakis, Pavithra Harsha Market forecasts show that e-commerce stands to ultimately inherit a significant proportion of the retail market. Gaining a competitive edge in this sector is of utmost importance to any firm’s success. The model presented in this work guarantees customer satisfaction by providing relevant recommendations at personalized prices in a way that balances profit maximization with business operations. We demonstrate its value on actual e-tailer data. 2 - Pricing for a Satellite Service Provider Charles Thraves, MIT, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States of America, thraves@mit.edu, Georgia Perakis We present a pricing optimization formulation of the data plans of a satellite service provider. First, to estimate reservation prices for all customers (including unobserved customers) we deal with the missing data problem. We introduce a MIP formulation and develop properties and heuristics for the problem. We develop analytical bounds for our heuristics and conclude that they can help the company increase its profits by more than 10%. 3 - Scheduling Promotion Vehicles to Boost Profits Lennart Baardman, MIT, Operations Research Center, Cambridge, MA, 02139, United States of America, baardman@mit.edu, Kiran Panchamgam, Danny Segev, Georgia Perakis, Maxime Cohen Retailers use promotion vehicles (e.g. flyers, commercials) to increase profits. We model how to assign promotion vehicles to maximize profits as an NLIP. The problem is NP-hard and even hard to approximate. However, we construct an epsilon-approximation in the form of an IP of polynomial size. Also, we propose a greedy algorithm with a provable guarantee and on average near-optimal performance. Finally, using supermarket data we show that our model can lead to a significant increase in profits. 4 - Dynamic Pricing through Combinatorial Methods Jeremy Kalas, MIT, Cambridge, United States of America, jkalas@mit.edu, Swati Gupta, Kiran Panchamgam, Georgia Perakis, Maxime Cohen We explore fast combinatorial methods for the multi-period, multi-item Promotion Optimization Problem under general demand functions that depend on past prices. As the problem is NP-hard for large memory, we consider an approximation via the reference price model, and give a PTAS. We extend this model to handle cross-item effects using a ``virtual” reference price. We report a projected 4-6% increase in profits on real-world data sets, in collaboration with the Oracle Retail Science group. SD46 46-Room 104A, CC Empirical Operations Management: Services Sponsor: Manufacturing & Service Oper Mgmt/Service Operations Sponsored Session Chair: Robert Bray, robertlbray@gmail.com 1 - Free Riding and Auto Recalls: An Asymmetric Dynamic Discrete Choice Game Ahmet Colak, Northwestern University, 1116 W Loyola Ave Apt 3S, Chicago, IL, 60626, United States of America, a- colak@kellogg.northwestern.edu, Robert Bray We structurally study auto recalls. Two agents can initiate a recall: the manufacturer and the federal regulator. Initiating recalls is an entry game with optimal stopping. We unexpectedly find that the regulator has no deterrence power over the manufacturer, and that the two agents free ride off each other’s recall efforts. Free riding decreases the regulator’s and manufacturer’s median recall probabilities by 2.9% and 0.7% respectively, and increases society’s exposure to defective parts.

4 - Analysis of Competitive Pricing with Multiple Overlapping Competing Bids in Revenue Management Goutam Dutta, Professor, Indian Institute of Management, Wing 3, PMQ Area, Roon No 3H, Old Campus, Ahmedabad, 380015, goutam@iimahd.ernet.in, Sumeetha Natesan We analyze the competitive pricing situation of one company with more than one competitor. Based on the past experience one can have some idea about what the competitors will bid which can be described by various distributions. The prices of company and its competitors follow independent, overlapping uniform distribution. First we derive expression for probability of win and then, we attempt to derive conditions for maximizing the expected contribution to profit. 5 - Data Science, Operations Research, Analytics and Revenue Management Jon Higbie, Managing Partner & Chief Scientist, Revenue Analytics, 3100 Cumberland Blvd., Suite 1000, Atlanta, GA, 30339, United States of America, jhigbie@revenueanalytics.com The term Data Science and the job title Data Scientist are much in vogue right now. How does Data Science relate to Operations Research? What lessons might we learn about the past boom in Operations Research that can be applied to the current boom in Analytics? How is the increased emphasis on Business Analytics impacting Revenue Management? These questions will be explored through a series of case studies, and discussion. SD44 44-Room 103B, CC Algorithmic Revenue Management with Strategic Customers Sponsor: Revenue Management and Pricing Sponsored Session Chair: Vivek Farias, Associate Professor, MIT, 100 Main Street, Cambridge, United States of America, vivekf@mit.edu Co-Chair: Yiwei Chen, ywchen@ckgsb.edu.cn 1 - Analysis of Discrete Choice Models: A Welfare-Based Framework Guiyun Feng, University of Minnesota, 1006 27th Avenue SE,, Minneapolis, MN, 55414, United States of America, fengx421@umn.edu, Zizhuo Wang, Xiaobo Li We propose a framework for discrete choice models through a welfare function. The framework provides a new way of constructing choice models. It also provides great analysis convenience for establishing connections among existing choice models. We define a new property in choice models: substitutability/complementarity and study conditions for a choice model to be substitutable. We show that our framework is flexible in this property, which is desirable in capturing practical choice patterns. 2 - Robust Dynamic Pricing with Strategic Customers Yiwei Chen, Assistant Professor, Renmin University of China, No. 59 Zhongguancun Street, Beijing, China, chenyiwei@rbs.org.cn, Vivek Farias We consider the canonical revenue management problem that a seller sells finite number of a product over a finite horizon via dynamic pricing. We assume that customers are forward looking with heterogeneous strategic factors (time discount rates and monitoring costs). We propose a class of pricing policies that achieve at least 29% of revenue under an optimal dynamic mechanism. Our policies require no knowledge of customers’ strategic factors and ensure customers to behave myopically. 3 - Managing Multi-period Production Systems with Limited Process Flexibility Yehua Wei, Fuqua School of Business, Duke University, 100 Fuqua Drive, Durham, NC, 27708, United States of America, yehua.wei@duke.edu, Cong Shi, Yuan Zhong We develop a theory for the design of process flexibility in a multi-period production system. We propose and formalize a notion of “effective chaining” termed the Generalized Chaining Condition (GCC). We show that any partial flexibility structure that satisfies GCC is near-optimal under a class of policies called the Max-Weight policies. Furthermore, we show that GCC can be satisfied using just k arcs, where k is the equal to the number of products plus the number of plants. 4 - Optimal Dynamic Pricing with Patient Customers Yan Liu, University of Minnesota, Room L123, ME Building, 111 Church ST, Minneapolis, MN, 55455, United States of America, liux0984@umn.edu, William Cooper We consider a single-product pricing problem in which a fraction of customers is patient and the remaining fraction is myopic. A patient customer will wait up to some fixed number of time periods for the price to fall below his valuation, at which point the customer will make a purchase. If the price does not fall below a patient customer’s valuation at any time during those periods, then the customer will leave without buying. We identify the structure of an optimal pricing policy.

134